Observations of aerosols in the free troposphere and marine boundary layer of the subtropical Northeast Atlantic: discussion of processes determining their size distribution

Observations of aerosols in the free troposphere and marine

boundary layer of the subtropical Northeast Atlantic:

Discussion of processes

determining their size distribution

Frank

Raes

• Rita

Van

Dingenen

• Emilio

Cuevas

2, Peter

F. J. Van Velthoven

3 and

Joseph

M. Prospero

4

Abstract. During July 1994, submicron aerosol size distributions were measured at two sites on Tenerife, Canary Islands. One station was located in the free troposphere (FT), the other in the marine boundary layer (MBL). Transport toward these two sites was strongly decoupled: the FT was first affected by dust and sulfate-laden air masses advecting from North Africa and later by clean air masses originating over the North Atlantic, whereas the MBL was always subject to the noaheasterly trade wind circulation. In the FT the submicron aerosol distribution was predominantly monomodal with a geometric mean diameter of 120 nm and 55 nm during dusty and clean conditions, respectively. The relatively sma•11 diameter during the clean

conditions

indicates

that the aerosol

originated

in the upper troposphere

rather

than over

continental

areas

or in the lower stratosphere.

During dusty

conditions

the physical

and

measurements. The MBL aerosol size distribution was bimodal, but the relative concentration

of Aitken and accumulation mode varied strongly. The accumulation mode can be related to

cloud processing

of the Aitken mode but also to pollution

aerosol

which was advected

within

the MBL or entrained from the FT. No bursts of nucleation were observed within the MBL.

1. Introduction

Subtropical oceanic atmospheres, adjacent to the western parts of the continents, are characterized by a distinct temperature inversion. This is a consequence of heating by compression of the subsiding air over a cold upwelling ocean. The inversion tops the marine boundary layer (MBL) at altitudes that range between 500 m, close to the center of the subtropical high, and 2000 m in the trade-wind region closer to the tropics. When the condensation level is lower than the inversion level, water vapor will condense on the available aerosol particles to form stratiform clouds. Because of the large scale of subtropical subsidence, the MBL stratiform clouds can extend over very large areas and cover 30% of the

1Environment Institute, Joint Research Centre, European Commission, Ispra, Italy.

:Instituto

Nacional

de Meteorologia,

Tenerife,

Islas Canarias,

Spain.

'•Royal

Netherlands

Meteorological

Institute,

De Bilt,

Netherlands.

*Division

of Marine

and Atmospheric

Chemistry,

University

of

Copyright 1997 by the American Geophysical Union.

Paper number 97JD01122.

0148-0227/97/97JD-01122509.00

world's oceans [h•ten•ational Satellite Cloud Climatology Project (ISCCP), 1992]. Charlson et al. [ 1987, 1992] drew attention to the possible role of these clouds in climate regulation and climate change. Assuming a constant liquid

water content, these authors estimated that a 30% change in

the cloud droplet number concentration of MBL stratus clouds

could

lead

to a global-mean

radiative

forcing

of-2 W/m

2.

Cloud droplets form on a fraction of the aerosol particles, called cloud condensation nuclei (CCN). The estimates by

Charlson et al. have therefore focused a lot of aerosol research

on the origin and behavior of MBL aerosols and CCN. Even for the simplest case of the clean MBL, where sea salt and sulfate derived from dimethylsulfide (DMS) are the major aerosol constituents, there is not yet an agreement regarding the following two issues: (1) What is the relative contribution of sea salt and DMS-derived particles to the total number

concentration of MBL CCN? (2) Is the number of DMS-

derived aerosol particles controlled by nucleation of condensable species within the MBL, or are they formed in the free troposphere (FT) and subsequently entrained into the MBL? In case of a polluted MBL, aerosol particles and CCN are provided by advection from the continents. Here the following question may be raised: (3) What is the relative importance of transport within the MBL, and long-range transport in the FF with subsequent entrainment, as a way of perturbing the MBL and its clouds? All these questions have

to do with how to describe aerosol and, eventually, cloud

droplet number concentrations in climate models; whether to focus more on transport issues, on atmospheric chemistry and

aerosol dynamics, or on primary aerosol sources.

A quantitative answer to the questions will eventually

require Lagrangian studies in order to untangle aerosol

processes from transport processes, as well as long-term

measurements yielding meaningful statistics about the type of conditions or transport patterns that are prevailing. However, simultaneous characterization of the aerosol in the MBL and in the overlying FF should also give insights in the processes

that are possibly involved. With this in mind, we have performed measurements on Tenerife, an island located in the NE subtropical Atlantic, where we operated two stations, one

in the MBL and one in the free troposphere (FF). During

summer, both stations are subject to clean maritime conditions

and to outbreaks of anthropogenic aerosols from Europe and

mineral dust from northern Africa.

In this paper we focus on measurements of the size

distribution of submicron aerosols and attempt to explain the observed variability (and regularities) in terms of aerosol dynamic processes and transport patterns.

2. Sites and Meteorology

Measurements were performed during July 1994, at two

sites on Tenerife (Canary Islands, 16 ø 30' W, 28 ø 18'N ):

Punta del Hidalgo (PDH), (50 m above sea level (asl)) and Izafia (IZO), (2360 m asl).

Punta del Hidalgo (PDH)

Measurements were taken from the top of a 50 m high

lighthouse. It is located on the shore of the NE extremity of the

island and is directly exposed to the northeasterly trade winds. PDH is a new site and no climatology of trajectories exists. Calculations of 5 days back trajectories during July 1994 (see Figure l a), 1995 and 1996 show that during this month the site is usually impacted by air masses subsiding around the high of the Azores. Often, these air masses will pass close to the Iberian Peninsula or even have their origin over western and central Europe, transporting enhanced levels of anthropogenic pollution through the MBL to Tenerife. On 1 out of 90 days, air masses in the boundary layer came straight

from Africa.

Izafia (IZO)

Measurements were taken at the Instituto Nacional de

Meteorologia (INM) - Global Atmosphere Watch (GAW) Observatory, which is located at 2360 asl, on the mountain

ridge that crosses the NE part of the island. Climatologies of

back trajectories [Sancho et al., 1992, Merrill. 1994, Cuevas et al., 1996] show that in summer the site is mainly impacted

by air subsiding from the •nid and upper troposphere over the

North Atlantic and by continental air masses from northern Africa. Frequently, the latter have their origin over the Mediterranean or central Europe.

Radio soundings from Santa Cruz de Tenerife show that during July the subsidence temperature inversion is located between 1000 and 1500 m. This inversion separates the MBL air masses arriving at PDH from the FF air masses arriving at IZO. However, during daytime, when the (volcanic, dark) island is heated, upslope winds will bring air

from lower levels and possibly from below the inversion to IZO. Urban development and industry along the coast of the island as well as biogenic emission from the forests on the mountain up to 1900 m strongly alter the composition of the upslope winds, compared to what is sampled at PDH.

For the period discussed in this paper (July 7-18, 1994), back trajectories were calculated with a variety of models: two three-dimensional (3-D) models based on the European Centre for Medium Range Weather Forecast (ECMWF) and the National Meteorological Center (NMC) wind fields respectively, an isobaric model based on the United Kingdom Meteorological Office (UKMO) model and isentropic trajectories based on NMC wind and thermodynamic data. They all show the same general picture, however relevant differences also exist. Here we will show only the Royal

Netherlands Meteorological Institute (KNMI) trajectories

(Figure 1). They are based on ECMWF wind fields, which have a horizontal resolution of 0.5 ø, and are available every 6 hours. The relatively good performance of the ECMWF model over ocean regions has been demonstrated by Fuelberg et al. [1996]. The sensitivity of the KNMI trajectories to time and spatial resolution showed that a 6 hour, respectively, 1 o x 1 o resolution is largely sufficient to describe 3-D air mass trajectories in synoptic systems [Scheele et al., 1996].

In Punta del Hidalgo, air masses were mainly from the open Atlantic, except between July 11 (0600) and July 13

(0600) when they traveled closer, or according to the isobaric UKMO model, over the Iberian coast (F. McGovern, private

communication, 1997; see also Figure la). Izafia, on the

contrary was influenced by European-African air between July

8 and 14, followed by midtropospheric Atlantic air between July 15 and 18 (see Figure lb and lc). This analysis of the origin of air masses is confirmed by the chemical composition and loadings of the aerosol: mainly clean air at PDH and dust followed by clean air at IZO (F. McGovern, private communication, 1997). In this paper we will focus'on the size

distribution of the submicron aerosol, and collate them with

the transport patterns and aerosol processes.

3. Instrumentation

At both sites, aerosol size distributions were obtained with

identical differential mobility particle sizers (TSI DMPS model 3932) operating in the size range between 16 and 620 nm and giving a distribution every 20 min. The two systems were intercompared side by side using laboratory aerosol. The number concentrations resulting from integrating the number

distribution were equal within 10%, and the geometric mean diameters of the distribution were within 5%. The DMPS at PDH failed after July 17.

Apart from the TSI DMPS, a second system was used at

IZO, consisting of a short and medium length Vienna-type differential mobility analyzer (DMA), [Winkhnayer, 1987]

operated in parallel. The short, ultrafine DMA (UFDMA) was operated with a TSI 3025A particle counter, while the medium DMA (MEDMA) used a TSI 3010 counter. The size ranges scanned by he UFDMA and MEDMA were 3.5-20 nm

and 20-620 nm, respectively. Both scans and hence the

a 3-D 5-day backward trajectories (KNMI/ECMWF) ending at 975 hPa

first begindate: SATURDAY 2 JULY 1994 12 GMT last enddate: SATURDAY 16 JULY 1994 t2 GMT

70øW 60øW 50øW 40øW 30øW 20øW 10øW 0 ø 10øE

o ,>. ,[15] 705 .-,.

'":':"

' 0]

659 8]

807

16

7:: .../.. 5ON , '

50øW 40øW 30øW 20øW 10øW

6O0

7OO

8OO

9OO

lOOO -12o

3-D 5-day backward trajectories ending on 7-16 July 12Z at 975 hPa

i i i i i i i

7 July

8 July .... 9 July ...

10 July 11 July 12 July - - - 13 July ... _ 14 July ... 15 July -- 16 July

- _

)•:-..

"'" ....

_

I I I I I I i I i

-108 -96 -84 -72 -60 -48 -36 -24 -12 0

T•me (h) before 940707 12 Z

Figure la. Compilation

of 5 days

back

trajectories

ending

at Tenerife,

at 975 hPa.

(Punta

del Hidalgo

(PDH)), at 1200

UTC between

July

7 and

16 1994.

(a) Horizontal

projection:

the

numbers

[xx]

yyy

given

at the

beginning

of each

trajectory

indicate

the

end

day

(xx) and

the

pressure

level

(yyy)

where

they

start.

(b) vertical

projection.

times and size classes where both IZO systems yielded valid data. they are presented in what follows as averages from the

two systems. In other cases, data from one system are taken

into account. At PDH, size distributions were measured at 40- 50% RH, whereas at IZO the RH was always lower than 25%. These relative humidities were obtained from those recorded outside, taking into account the higher temperature in the instrumentation room. In what follows the shown size distributions can therefore be considered as those of the aerosol in the dry state.

Other measurements that will be used for interpreting the

size distribution measurements are as follows: (1) at IZO:

surface ozone, measured with a Dasibi UV absorption instrument, CO2, with a nondispersive infrared Ultramat-2 analyzer, dry and wet temperature, and global radiation with a

Kipp and Zonen pyranometer; (2) vertical information of

temperature, humidity, and wind over Tenerife, from sea level

to 30 km height, from rawinsondes,

released

twice a day

Very High Resolution

Radiometer

(NOAA/AVHRR) images

4. Observations

Cloudiness From AVHRR Images

The images are taken at the receiving station of Dundee

a 3-D 5-day backward trajectories (KNMI/ECMWF) ending at 750 hPa first begindate: SATURDAY 2 JULY 1994 12 GMT last enddate' WEDNESDAY 13 JULY 1994 12 GMT

30ow 20øW 10øW 0 ø 10øE 20øE

20øW 10øW 0 ø 10øE

60O

7O0

8OO

9OO

1000

3-D 5-day backward trajectories ending on 7-13 July 12Z at 750 hPa

,

,\ i i i i i i i i i

\,.,

• July

--

".... • ..." .... '... 9 July ... '-- .... • j • ...' ... 10 July ...'"... /.,-' '%.,..•.,,,' • ... . 11 July .... ,,,' .... /'... '.. '"•'•..---•'"'•' /- ... ,, ':< .... , 12 July ---

,; //! ", • x ' .- '

,.,./ ', '., • _ - -.. .'

.,•. ', '. '. _ _ -

: ...

7:::::-'-:'"

"

.

ß ß o. o. ..

20 -108 -96 -84 -72 -60 -48 -36 -24 - 2 Time (h) before 940707 12 Z

Figure

lb. Compilation

of 5 days

back

trajectories

ending

at Tenerife,

at 750 hPa

(Izafia

Station

(IZO)),

at 1200

UTC

between

July

7 and

13 1994,

i.e., when

dust

was

observed

at IZO. (a) Horizontal

projection,

(b) vertical

projection.

Atlantic

and western

Europe,

and include

the Canary

Islands

period

were collected

and can be seen

at the WWW site

http://rea.ei.jrc.it/--vandinge/sat94/sat94.htrnl.

By visually

subtracting

channel

2 (VIS) and channel

4 (IR) images,

a

rough

impression

of low-level

cloudiness

can be obtained.

Detailed

analysis

of these

satellite

data

is outside

the scope

of

this study.

During

the period

discussed

here,

the cloudiness

over

the

between

the Canary

Islands

and the coast

of Portugal,

i.e., the

area crossed

by the MBL air parcels

during

the last 36-48

hours

before

reaching

Tenerife.

This area

is generally

free of

level clouds

are present

immediately

upwind

of Tenerife

and

from July 13 onward

the cloudiness

becomes

complex,

with

often high-level,

probably

convective

clouds

near and over

Portugal.

Structure of the Atmosphere above Tenerife

Plate la shows, as a 2-D contour plot, the time history of

the vertical

profile

of the dew point,

obtained

from radio

soundings

at Santa

Cruz de Tenerife.

Santa

Cruz is located

at

the trade winds. The first 600 m of the radio soundings are

a 3-D 5-day backward trajectories (KNMI/ECMWF) ending at 750 hPa first begindate: SATURDAY 9 JULY 1994 12 GMT

last enddate: MONDAY 18 JULY 1994 12 GMT

70øW 60 øw 50øW 40øW 30øW 20øW 10øW 0 ø 10øE 20øE

ß

.. "'::¾:11

'•..;.,.:.x,:•

,..:.

...

!•.1....4.•..!•

...

;,

...

::','-'"

:.. - ...

:

... ß

...

ß

...

...

)':'::

...

.,

...

/.

...

/4•o,

'•i

...

:•::

...

•-- '

.... .

...

50øW 40øW 30 TM 20øW 10øW

5OO

600

700

8OO

9OO

1000

3-D 5-day backward trajectories ending on 14-18 July 12 UTC at 750 hPa

I I I I I I I I I

July July 18 July

-."'?..-':::-::

...

<:...

ß

_

_

-120 -108 -96 -84 -72 -60 -48 -36 -24 -12 0 Time (h) before 940714 12 Z

Figure

lc. Compilation

of 5 days

back

trajectories

ending

at Tenerife,

at 750

hPa

(IZO),

at 1200

UTC

between

July

14

and

18 1994,

i.e.

when

clean

air

was

observed

at IZO. (a) Horizontal

projection,

(b) vertical

projection.

large variability in time and with altitude. The station is

impacted

by intrusions

of dry upper

tropospheric

air on the

night

between

July 10 and 11 and more

clearly

between

July

14 and 17. During these

intrusions

a distinct

layering

is

observed,

and

IZO is representative

only

for a relaiively

small

portion

of the free troposphere.

During the other periods,

free

troposphere

is more

homogeneous,

at least

with respect

to

its thermodynamic properties.

FT Size Distributions at lzafia (Nighttime)

A good

picture

of the evolution

and variability

of the size

distributions

is given by the 2-D contour

plot (Plate lb),

showing

dN/dlog(Dt,

) as function

of Dt• and time, for the

whole measuring period. The strong diurnal variations, with

high concentrations

of small particles

during

daytime

stand

out clearly. We will first consider the nighttime

measurements,

because

they are least disturbed

by upslope

shape

of the corresponding

size distributions

is given in

Figures

2a and

2b, as a frequency

plot

of dN/dlog(Dt,)

versus

Dr,. The averages of the modal parameters of the distributions are given in Table 1.

A first observation

is that Fr size distributions

are

predominantly

monomodal.

This

can

be explained

by the

absence

of cloud

processing

and

the long

residence

times

in

the Fr, during

which

coagulation

will reduce

any

initial

Dew point and size distributions in IZO and PDH

7

8

9

10 11 12 13 14 15 16 17 18 19 20Dew

point

8000 • "• - .• • •. ' [C]

6000

•

<>

5

o -10

.

.

. -25

-35

-- •5

•' •

.. ,• •

-•---•

-50

....

• ----• ... ----• ... •--•---•'• -100 •

•-I •-I! ,

100 .. •

F

lO

_

_

_

-

100 -

-- _ --

--

--

lO

, , I , , , I , !

7 8 9

,,I:,,I::,

I I I I I ' , I , , , I : ! I i I I I I I | i

10 11 12 13 14 15 16

day of July

I I I I I I I I , , , I , I

17 18 19 20

cN/dlogDp

2500 2OO0 1500 1000 900 8OO 7O0

6OO

5O0

40O

3O0 25O 2O0 100

50 0

Plate 1. (a) Time history

of the vertical

dew point

profile,

obtained

from radio soundings

released

twice a day at Santa

Cruz de Tenerife. The horizontal line at 2360 m indicates the location of the Izafia Station (IZO) (b) Time history of the submicron size distribution measured at Izafia (IZO). The size distributions switch between unimodal during nighttime and

bimodal

during

daytime.

Two episodes

can be discerned:

the nighttime

distributions

change

abruptly

on July 14 (0000),

which coincides

with the arrival

of dry upper

tropospheric

air at the station.

(c) Time history

of the submicron

size

distribution

at Punta

del Hidalgo

(PDH) 50 m above

sea level. The size distribution

is always

bimodal,

with strong

processing in the subtropical FT is supported by the ISCCP data, which show that the frequency of medium and high clouds in this area is typically < 10%. It is worth noticing, however, that according to theory, coagulation eventually results in lognormal distributions with a geometric standard deviation of 1.4 [Lai et al., 1972], whereas coagulation plus

condensation would result in even narrower size distributions. Here the observed value is around 1.9!

As mentioned before, all back trajectory models show that before July 14 the air masses arriving at IZO passed over the Sahara and, with some exceptions, over the Mediterranean before passing the Sahara. During the course of July 14, the transport pattern changed and by July 15 trajectories originated in the middle atmosphere above the North Atlantic. Contrary to this gradual change suggested by the trajectories and by the dewpoint (Figure 2a), the composition of tile atmosphere changed more abruptly at July 14 0000. The implication of this change from dusty to clean conditions is a dramatic decrease of the geometric mean diameter of the

submicron aerosol from l l7 nm to 55 rim. We now discuss

the dusty and clean episodes in more detail.

Episode FT-I. July 7, 0000 to July 14, 0000 (Figure 2a). The average diameter of the aerosol is fairly constant (ll 7 + 13 rim, see Table l), but the number concentration is more variable. This variability can be related to the large variability

in horizontal and vertical transport over northern Africa

(Figure lb). d'AImeida and Schutz. [1983] report that during

desert dust storms particles in the range 0.1-0.2 lure Dr, can

reach

number

concentrations

up to 1500

cm

-•. It is theretore

possible

that

tile observed

400-600

cm

-• submicron

particles

at

this sulfate mass is in the submicron fraction, and that the

Ca++/SO4 -- ratio is smaller in the submicron than in tile supermicron fraction, suggesting different origins tbr these fi'actions (D. Savoie and J.P. Putaud, personal communications. 1997), (3) during the episode discussed here, the regression of nighttime (12 hour averages)

submicron

aerosol

volume

on sulfate

yields r 2 = 0.84,

submicron fraction.

The decrease of the number concentration and appearance of smaller particles at July I I 0000, coincides with drier air

around Tenetire (see Plate i a), which is likely to be a

remainder of air transported from higher altitudes, as is shown by the back trajectory ending on July I I 1200. Such downward transport was more clear after July 14.

Episode FT-II, .July 14, 0000 to .July 18, 1200 (Figure 2b). The sudden decrease in the particle size coincides with a decrease in water vapor and a strong increase with 03 (see Figure 3c). Plate l a shows that a layer of extremely dry air is subsiding over Tenefife, and the trajectories show that this layer has its origin in the middle troposphere. The monomodality and relatively small average diameter of the

aerosol distribution show that the aerosol is relatively fresh

and formed by nucleation and subsequent condensation and coagulation higher up in the troposphere. An aged aerosol

Izana, 2350 m asl !zana, 2350 m asl

lOOOO

i ...

_

... ,

...

loooo

I ...

Diurnal Variation of the Aerosol at IZO

The most prominent feature of the aerosol at IZO is the diurnal variation in the number concentration and size distribution of the aerosol. This diurnal variation in particle

concentration has been noticed before. In fact, diurnal

variations have been documented for most of the atmospheric constituents measured at IZO. They are explained by upslope winds that evolve during daytime and that bring air from

lower levels and probably from within the MBL to the

observatory. It has also been noticed before that during dusty

conditions the upslope winds are weaker, and IZO is less

influenced

by air at lower levels [Cuevas

et al., 1991].

analyses. For what concerns aerosol particles, our

measurements show for the first time that the diurnal variation

in number concentration is predominantly due to an increase

in the ultrafine

particles

(Plate lb, Figure 3d). During clean

air near

IZO, since

particles

were

detected

down

at the lower

nucleation

seems to have happened

and subsequently

quenched

in the upslope

air before

it arrived

at IZO, since

the

smallest

particles

detected

were significantly

larger

than in

clean conditions. This difference can be explained by the

known slower upslope

transport

during dusty conditions,

and/or

by quenching

of nucleation

when

the upslope

air mixes

o 04

364

562- C02 (ppmV) -

560

558

556

554

_ _

552 • • i

o •4 :lb)

012

010

0 08

OO6

ø188

5 8 9 10 11 1;2 1.5 14 15 16 17 18 19

so -(c)

i , t•'/

- ' 'v.-• i

o -. ... ... , ... , ... , .... , .... , ...

-(d) Rod,or,on ond CN (•/cmS) smolle•. tho%,

7 8 9 10 11 12 13 14 15 16 17 18 19 20

Do• of July

Figure 3. Diurnal variations of a variety of parameters measured at IZO observatory, documenting the upslope transport of air during daytime: (a) half hourly CO2 values (ppmv) and (b) corresponding standard deviation (%), (c) ozone and water vapor, (d) global radiation and aerosol particles smaller than 10 nm.

Although the observed nucleation is most likely a local phenomenon, induced by the photochemistry in the perturbed air masses reaching IZO, it must be looked at with care since it might mask in situ nucleation over large areas in the undisturbed free troposphere at the altitude of IZO. There is

no evidence in our data set that the latter is the case. The

ultrafine particles never appeared before humidity, ozone or

CO2 signaled the arrival of upslope winds at IZO, and they usually disappeared abruptly after the sun sets and the upslope wind collapses. The latter is true except on the evenings of

July 16 and July 17 (see Plate lb). During these nights the winds at IZO drop below 2 m/s, and the back trajectories show local transport near Tenerife. It is therefore likely that the air that has been pumped up during the previous days is lingering on during nighttime.

MBL Size Distributions at Punta del Hidalgo

Plate l c shows the time history of the aerosol size distributions at Punta del Hidalgo, and Figures 2c, 2d, 2e and 2f the corresponding distributions as frequency plots. Table 2 gives the average modal parameters of the aerosol.

An immediate observation is that the MBL size distribution

is bimodal most of the time, showing an Aitken mode around

60 nm Dp and an accumulation mode around 200 nm Dr,.

High concentrations of ultrafine particles are not apparent, showing that bursts of nucleation did not occur in the MBL

upwind of and close to PDH. This lack of nucleation bursts

raises the problem of the origin of the Aitken mode particles,

which we will discuss later. The accumulation mode is commonly explained by the cycling and chemical' processing of the Aitken mode aerosols in the stratiform cloud deck

[Hoppel et al.. 1986, 1994]. However this explanation does not seem to apply to the accumulation modes observed before July 12. when AVHRR images do not show obvious low-level

clouds upwind of Tenerife. This and the generally large

variation in the relative importance of Aitken and accumulation mode asks for a closer look into the origin and

behavior of these modes. We have used the KNMI model to

calculate 5 days back trajectories ending in the MBL near

Tenerife at 900, 925, 950. and 975 hPa. In the discussion

below we will only refer to the 925 and 975 trajectories. The trajectory analysis and the patterns seen in Plate l c allow identification of four episodes.

Episode MBL-I, July 9, 1800 to July 11, 0600 (Figure 2c). The distributions are characterized by a sharp Aitken mode around 60-70 nm. The 925 and 975 trajectories ending on July 10 1200 originated five days earlier in the middle troposphere at 560 and 659 hPa, respectively. They reached

the inversion (taken as the 900 hPa level) at about 40øN. 24

to 36 hours prior to arrival in PDH. On the basis of these trajectories an average subsidence velocity of 0.8-1.0 cm/s can be inferred, which is within the range of entrainment velocities determined during the Atlantic Stratocumulus Transition Experiment (ASTEX): i.e., 0.3-1.0 cm/s [Bretherton et al.. 1995]. It is therefore likely that FI' air

entered and mixed with the MBL without much resistance.

The most straightforward explanation for the MBL Aitken mode is therefore that it originated from the preexisting FT aerosol that entered the MBL and evolved by condensational growth. Condensational growth would increase the size of the

Table 2. Average MBL Aerosol Properties, Based on Fits of Two Lognormal Distributions to the Measured Distributions

Aitken Mode Accumulation Mode

N, D,,.• . N. D,,:,

Episode Type cm -• STP nm c•e cm -3 STP nm c• e

MBL-I mixed 319 (152) 66 (9) 1.38 (0.16) 349 (115) 120 (40) 2.00 (0.28) MBL-II mixed 447 (327) 65 (7) 1.41 (0.12) 325 (73) 217 (14) 1.38 (0.08) MBL-11I clean 321 (169) 52 (6) 1.47 (0,09) 207 (51) 209 (16) 1.35 (0.07) MBL-IV clean 228 (44) 52 (2) 1.40 (0.05) 249 (23) 178 (5) 1.34 (0.02)

MBL, marine boundary layer

Values in parentheses' I standard deviation.

length of the condensing molecules, as is the case here [Friedlander; 1977, p. 253]. This would explain the significantly lower value of the geometric standard deviation of the MBL Aitken mode (1.38 + 0.16) compared to that of a

clean FF aerosol (1.89 + 0.1). We cannot exclude nucleation

occurring upwind of Tenerife, during the initial mixing of dry FT air with humid MBL air [e.g. Hegg et al., 1992] and that the freshly nucleated aerosol evolved and contributed to the Aitken mode. Apart from the sharp Aitken mode, a broad accumulation mode is also present (see Figure 2c). Cloud processing is not a viable explanation, since no low-level clouds were present between the point where the trajectories entered the MBL and PDH. The FT aloft, however, was heavily laden with aerosol, which we interpreted as anthropogenic sulfate. The AVHRR pictures and back trajectory analysis show that this FF aerosol extends NE of Tenerife toward Portugal. It is thus possible that entrainment has partially flushed the clean MBL with the aerosol from the F-F, and that the latter is at the origin of the observed accumulation mode. This scenario is supported by the correspondence between the geometric mean diameter and standard deviation of the MBL accumulation mode particles and the FT aerosol aloft. A narrowing of the size distribution

is indeed not expected since the particles are now larger than the mean free path length. The scenario is also supported by

the chemical analysis of the MBL aerosol, which shows traces of mineral dust (D. Savoie, personal communication, 1996).

Episode MBL-II, July 11, 0600 to July 13, 0600 (Figure 2d). Although low-level clouds are still not clearly present upwind of PDH, the aerosol distributions are characterized by prominent Aitken and accumulation modes. The 925

trajectory ending on July 11 1200 still originated at 580 hPa,

and could still explain the large Aitken mode. The 975 trajectory. on the other hand, remained below the 800 hPa level during the previous 5 days (see Figure 4). Although the latter trajectory remains over the ocean, the isobarit 1000 hPa trajectory of the UKMO model was located more eastward and passed over the Portuguese coast. This situation must have resulted in mixing of clean FT air with polluted BL air,

since black carbon and sulfate concentrations in PDH were

found to be higher during this period (McGovern, personal co•nmunication, 1997). Anthropogenic aerosols rapidly

develop an accumulation mode [e.g. Raes eta!., 1995], so the bimodal distribution during episode MBL-II can be explained

as the result of mixing of two air masses: one with a prominent Aitken and one with a prominent accumulation

mode. Both modes hide the contribution of FF aerosol entrained from aloft, which according to the aerosol chemical

analysis is still present. The trajectories ending on July 12 1200 show a similar pattern. During the afternoon of July 12 low clouds are present just upwind of Tenetire, so that the aerosol observed at PDH could have been minimally cloud

processed.

Episode MBL-III, July 13, 0600 to July 15, 0800

(Figure 2e).The distributions are characterized by prominent Aitken and accumulation modes. The 975 trajectories ending on July 13 and July 14 1200 stayed below the 850 hPa level all the time, whereas the 925 trajectories stayed below the 740 hPa level. The trajectories are passing farther away from the Iberian coast, and the chemical measurements show minimal anthropogenic impact. The air masses ending at PDH thus stayed predominantly within an unperturbed MBL for more than 5 days, and cloud processing has likely been the main

process shaping the bimodal size distribution. The short time variability in the Aitken mode must be explained by variability in the source of this mode. The variability in the accumulation mode is probably due to intermittent precipitation in the air mass. Hence episode MBL-III corresponds best to the classical picture of an undisturbed MBL, in which there is, on the average, a balance between the source of the Aitken mode particles, cloud processing to create the accumulation mode, and wet removal of the accumulation mode particles.

Episode MBL-IV, July 15, 0800 to July 16, 0600 (Figure 2f). There is again a prominent Aitken and accumulation mode, but the modal parameters changed slightly. This coincides with another change in mainly vertical transport. The 925 and 975 back trajectories ending on July 15 1200 originated again in the middle troposphere, at 520 and 700 hPa respectively, and enter the MBL 24- 36 hours upwind of Tenetire. The situation is very similar to Episode MBL-I, except for the clear presence of low level clouds along the trajectories toward PDH. It is most likely that cloud processing has eroded the aerosol entering the MBL to create the typical MBL Aitken and accumulation modes.

5. Discussion

FT

Mixing of dust with sulfate aerosols at IZO. The likelihood that the dust plume observed at IZO is mixed with

•o•w

3-D 5-day backward trajectories (KNMI/ECMWF) begindate: WEDNESDAY 6 JULY 1994 12 GMT

enddate' MONDAY 11 JULY 1994 12 GMT

50øW 40øW 30øW 20øW 10øW 0 ø

.

ß ..i...

30ON

--!,--

• .--')"

40øW 30øW 20øW

3-D 5-day backward trajectories ending at 940711 12 Z

4OO

500

600

700 8O0 900

1 ooo _,

!

900 hPa 925 hPa .... 950 hPa ... 975 hPa

20 -108 -96 -84 -72 -60 -48 -36 -24 -12 0

Time (h) before 940711 12 Z

Figure

4. Five-day

back

trajectories

ending

on July

11, 1200

at four

pressure

levels

within

the marine

boundary

layer

(MBL): 975,950,

925 and

900 hPa.

(a) horizontal

projections,

(b) vertical

projections.

They

show

that

the composition

of

the

MBL at PDH is made

up of two air masses,

one

of them

recently

originating

in the

free

troposphere.

internal/external mixing of these aerosols has implications for the radiative effect of the sulfate plume. We gave evidence for the fact that the sulfate particles observed during the dust event are predominantly in the submicron range, hence

externally

mixed

from the supermicron

dust

particles.

It is still

possible that the sulfate is internally mixed with the

submicron mineral dust aerosol that might be present.

External mixtures of submicron pollution aerosol and

supermicron

dust aerosols

have

been observed

by Clarke

et

They observed

a transition

from a predominantly

pollution

pollution/dust

aerosol

mixture

between

1200

and 2500 m, to a

predominantly

dust

aerosol

between

2500 and 4200 m. It is

not clear from their data whether the mixed aerosol layer

dioxide is still being oxidized. Such a plume is expected to interact more efficiently with dust, e.g., by heterogeneous

reactions or condensation. In this case the sulfate would be

moved into the supermicron size range, which is much less effective for scattering light [e.g. Dentenet et al., 1996]. It is therefore important to study this degree of mixing in more detail, for example, using single-particle chemical analysis.

High ozone concentrations at IZO. There is an ongoing discussion whether the high 03 levels at IZO, as seen for example on July 15, have an anthropogenic or natural origin [Schmitt, 1994: Prospero et al., 1995; Millan et al., 1996; Cuevas et al., 1996]. The size distribution observed during the clean episode FF-II might shed some light on this. The back trajectories and vertical profile of dew point during this episode support the upper tropospheric origin of the air mass

over IZO. Furthermore, the relatively small size of the aerosol

particles during nighttime argues for an aerosol that restilts from nucleation (the day betbre) in the upper troposphere rather than transport of aged anthropogenic or slratospheric

aerosol. However, this does not exclude yet that the ozone is

anthropogenic. Ozone and insoluble gaseous aerosol precursors may have been pumped up over industrial areas by convective clouds into the FT. while precipitation in such clouds removed most of the pollution aerosol [Prospero et al.,

1995]. However, one timber observation supports the natural origin of the air mass sampled at IZO, namely that the clean

FT aerosol size distributions do not show an accumulation

mode (see Figtire 2b). This observation could only be made consistent with convective transport of polluted air, if convective transport would remove with 100% efficiency the anthropogenic aerosol precursors and the pollution aerosol. This is unlikely. since convective transport is not necessarily connected with precipitation. We therefore conclude that during episode FT-II the air masses sampled at IZO during nighttime were not influenced by anthropogenic pollution. and that the ozone and aerosol had a natural origin.

MBL

There are a number of similarities between our

observations in the NE Atlantic, and observations made over

the remote Pacific Ocean. Covert et al. [1996] reported a dominant Aitken mode in areas of large-scale subsidence and during subsidence events behind cold fronts. Their Aitken mode is shifted to smaller sizes compared to what we observed during, for example, MBL-I. This is possibly due a faster downward transport (in particular behind cold fronts), hence shorter growth times for the newly formed aerosol. In

the Pacific trade-wind MBL, Covert et al., [1996] and Clarke

et al. [1996b] observed clearly distinct Aitken and accumulation modes, similar to what we observed during MBL-III. In all these cases it was the long residence and processing time within the MBL that resulted in the bimodal

distribution. This distribution can be considered the result of a

stationary state between a source of Aitken mode particles, cloud processing, and wet removal.

The fact that bursts of nucleation were not observed within the MBL is consistent with observations made elsewhere in

subtropical or trade-wind regimes [Clarke et at., 1996b, Clarke et al. 1996c, Covert et al. 1996. R. Van Dingenen, personal communication, 1995). These observations have risen the question about the origin of the Aitken mode particles observed in the MBL. Several hypotheses have been

presented in the literature, which can be divided into

"entrainment from the FF" [Raes et at., 1993; Clarke, 1993; Raes, 1995; Clarke et at., 1996b,c] and "nucleation within the

MBL" [e.g. Raes and Van Dingenen, 1992, Lin and

Chameides, 1993; Kerminen and Wexler, 1994; Pandis et at.,

1994; Coffinan and Hegg, 1995].

Entrainment. Our observations during episodes MBL-I, - II. and -IV, in particular the back trajectories originating in

the FT and the similarities between the MBL Aitken mode

and the FF aerosol, argue for a FF origin bf the MBL Aitken mode. During these episodes the transport from FT to MBL

seemed to occur as an injection of FF aerosol into the MBL at

a certain place. This is much different from the way Raes

[1995] modeled entrainment. i.e., as a constant flux

everywhere across the inversion. The latter type of entrainment can still be invoked for explaining the Aitken mode during episode MBL-III. The observed variability in the Aitken mode

number concentration indicates however that the associated

flux of particles is also very variable.

Other support for entrainment stems from the observation that despite the obvious decoupling of transport in the MBL and FT, and the differences in cloudiness in both layers, the

total number concentrations of the MBL and FT aerosol are

not very different. In Figure 5 we have plotted the frequency

distributions of aerosol number concentrations at IZO and

PDH for the whole duration of the period. It is shown that the background aerosol number concentrations at PDH and IZO

were

about

equal

at 600 cm

-3

STP.

A simple

explanation

could

coagulation will reduce the total number concentrations below

1000

cm • after

several

days

of transport

away

from

the source

MBL nucleation. In absence of this in situ source of new

particles in the MBL, the MBL number concentration must eventually equal the FT number concentration. or be lower in case of wet removal within the MBL. According to aerosol

dynamics

calculations,

15 lum:/cm

3 of dry aerosol

surface

is

measurements, O'Dos,'d et al. 11996] derived that a dry

surface

area

of about

3 lum:/cm

3 would

be sufficient.

Episodes

FT-I and FT-II yield average dry FT aerosol surface areas of

22 and

4.4 [im:/cm

•, respectively,

with

minimum

values

of 10

and !.2 lum:/cm

3. Considering

that within the MBL,

size of the entrained FT aerosol, we can conclude that the

injected or continuously entrained FT aerosol was probably sufficient to quench in situ nucleation in the MBL. We note also that in case of a continuous entrainment (e.g., episode MBL-III) the time scale of diluting the MBL is about 2 days (assuming an MBL height of 1000 m and an entrainment velocity of 0.6 cm/s), which is less than the (at least) 5 days the back trajectories stayed within the MBL before reaching

PDH.

In-situ nucleation. Classical nucleation theory predicts nucleation to be critically dependent on the environmental parameters [e.g.. Jaecker-Voirol and Mirabel, 1988]. Aerosol

dynamic models based on this theory therefore predict the

18% '

16%

ii

14%' 12% 10%

8%

6% 4% 2%

o o o o o

CD 0 0 0 0

,• PDH ',

,[31ZO

•i

o o o o o o o o o o

o o o o o o o o o o ON concentration bin ( STP #/cm 3)

Figure 5. Frequency of measured values of total number concentration in Punta del Hidalgo (day and night) and Izafia (night only), showing that between July 9 and 17, the background number concentration in the free troposphere was

essentially

equal

and around

600 cm

-3 STP.

Number

concentrations

were

obtained

by integrating

the TSI DMPS size

distributions.

upwind of PDH, as this would have restilted in a more

variable mean diameter of the Aitken mode. There is,

however, experimental evidence for a less critical (nonclassical) nucleation process [Weber et at., 1996], which opens the possibility that nucleation occurs more continuously

at low rates. Such a nucleation mechanism cannot be

dismissed by our data. Applying a continuous low rate nucleation of the order of 2.10 -3 cm-'•s -• in our AERO2 model [Raes, 1995] can indeed lead to a mode that peaks around 65

nm D/,, but with a tail down to the size of the nucleating

particles. This is because at a lower nucleation rate but constant formation rate of condensable material, more condensable material is available per particle. Hence each particle will initially grow very fast but eventually pile up around -- 65 nm where its growth becomes diffusion limited. The existence of an Aitken mode tail into the nanometer range could not be ascertained by the instrumentation operated at PDH. Low concentrations of nanometer particles would. in fact, be difficult to detect with any DMA system. because of the low sensitivity of these systems in the sub-10 nm range, and approaches like those of Weber et al. [1996] are essential. Sea spray. During our measuring period. production of submicron particles by sea spray can be neglected as a source of aerosol number. In the area upwind of Tenerife the average

wind speed is around 5 m/s. and according to O'Don'd et al.

[1993], such wind speeds would contribute about 10 sea

spray

particles/cm

3 larger

than

100

nm D•,.

This

is about

3 to

10% of the observed number of accumulation mode particles. and it is unlikely that it would be more in the Aitken mode.

Van Dingenen et al. [1995] have interpreted the

accumulation mode of marine aerosols as cloud condensation

nuclei. Our measurements show that during episode MBL-II the accumulation mode very probably consisted of pollution derived aerosol. which contained enhanced levels of black carbon. Since carbon containing particles might not all be capable of being activated in clouds, the contribution of sea salt to actual CCN (rather than accumulation mode particles) might have been larger than 3-10%. Single-particle chemical

analysis of CCN is required to quantitatively address the

question of the contribution of sea salt to CCN.

6. Summary

The observations of the aerosol size distributions at IZO and PDH allowed to identify various episodes, which were not

always detectable with chemistry measurements only. For

instance, the chemical composition of the aerosol at PDH

distinguished only between the polluted episode MBL-II on

the one hand, and the clean periods before and after (McGovem, personal communication, 1997). Back

trajectories, radio soundings. and AVHRR images, were essential to interpret the observations in terms of the origin of the air mass. and to show that FT-MBL transport, cloud processing. and mixing of air masses were the main processes

shaping the aerosol distribution of the MBL aerosol. Our measurements do not give quantitative answers to the three questions mentioned in the introduction but yield the

tollowing information:

1. There is a close link between the variabili.ty of MBL

aerosol and the changes in vertical transport patterns. This supports that entrainment or injection of aerosols from the FT is a major source of MBL aerosol. The entrainment process and its effect on MBL aerosol dynamics is more complex than originally proposed by Raes [1995l. For instance. the variability in the number concentration of the MBL Aitken mode suggests that the flux of aerosols by entrainment is not

constant. This can be due to variations in the entrainment

velocity. which is related to the structure of the MBL and the turbulent forces, and/or it can be due to the layered structure of the b--T, which we observed.

2. When injection of FF aerosol occurred, it occurred at a rather fixed distance upwind of the MBL site. It cannot be excluded that at the moment of injection, when dry FT air is mixed with humid MBL air, new particles were formed by rapid nucleation, which subsequently grew and contributed to